Language Modeling for Information Retrieval

252 Pages·2003·4.151 MB·

by  John Lafferty, ChengXiang Zhai (auth.), W. Bruce Croft, John Lafferty (eds.)

Language Modeling far Information Retrieval
THE KLUWER INTERNATIONAL SERIES
ON INFORMATION RETRIEVAL
Series Editor:
W. Brnce Croft
University ofM assachusetts, Amherst
Also in the Series:
MULTIMEDIA INFORMATION RETRIEVAL: Content-Based Information Retrieval
from Large Text and Audio Databases, by Peter Schäuble; ISBN: 0-7923-9899-8
INFORMATION RETRIEVAL SYSTEMS: Theory and Implementation, by Gerald
Kowalski; ISBN: 0-7923-9926-9
CROSS-LANGUAGE INFORMATION RETRIEVAL, edited by Gregory Grefenstette;
ISBN: 0-7923-8122-X
TEXT RETRIEVAL AND FILTERING: Analytic Models of Performance, by Robert M. Losee;
ISBN: 0-7923-8177-7
INFORMATION RETRIEVAL: UNCERTAINTY AND LOGICS: Advanced Modelsfor the
Representation and Retrieval of Information, by Fabio Crestani, Mounia Lalmas, and
Comelis foost van Rijsbergen; ISBN: 0-7923-8302-8
DOCUMENT COMPUTING: Technologiesfor Managing Electronic Document Collections,
by Ross Wilkinson, Timothy Amold-Moore, Michael Fuller, Ron Sacks-Davis, fames
Thom, and fustin Zobel; ISBN: 0-7923-8357-5
AUTOMATIC INDEXING AND ABSTRACTING OF DOCUMENT TEXTS, by Marie
Francine Moens; ISBN 0-7923-7793-1
ADVANCES IN INFORMATIONAL RETRIEVAL: Recent Researchfrom the Center for Intel
ligent Information Retrieval, by W. Bruce Croft; ISBN 0-7923-7812-1
INFORMATION RETRIEVAL SYSTEMS: Theory and Implementation, Second Edition,
by Gerald f. Kowalski and Mark T. Maybury; ISBN: 0-7923-7924-1
PERSPECTIVES ON CONTENT-BASED MULTIMEDIA SYSTEMS, by fian Kang Wu;
Mohan S. Kankanhalli;foo-Hwee Lim;Dezhong Hong; ISBN: 0-7923-7944-6
MINING THE WORLD WIDE WEB: An Information SearchApproach, by George Chang,
Marcus f. Healey, fames A. M. McHugh, fason T. L. Wang; ISBN: 0-7923-7349-9
INTEGRATED REGION-BASED IMAGE RETRIEVAL, by fames Z. Wang;
ISBN: 0-7923-7350-2
TOPIC DETECTION AND TRACKING: Event-based Infonnation Organization,
edited by fames Allan; ISBN: 0-7923-7664-1
Language Modeling for
Information Retrieval
Edited by
w. Bruce Croft
University ofM assachusetts,
Amherst u.S.A.
and
John Lafferty
Camiege Mellon University,
Pittsburgh, U.SA
SPRINGER-SCIENCE+BUSINESS MEDIA, B.V.
A C.I.P. Catalogue record for this book is available from the Library of Congress.
ISBN 978-90-481-6263-5 ISBN 978-94-017-0171-6 (eBook)
DOI 10.1007/978-94-017-0171-6
Printed on acid-free paper
All Rights Reserved
© 2003 Springer Science+Business Media Dordrecht
Originally published by Kluwer Academic Publishers in 2003
No part of this work may be reproduced, stored in a retrieval system, or transmitted
in any form or by any means, electronic, mechanical, photocopying, microfilming, recording
or otherwise, without written permission from the Publisher, with the exception
of any material supplied specifically for the purpose of being entered
and executed on a computer system, for exclusive use by the purchaser of the work.
Contents
Preface IX
Contributing Authors xi
1
Probabilistic Relevance Models Based on Document and Query Generation 1
lohn Lafferty, ChengXiang Zhai
1 Introduction 1
2 Generative Relevance Models 2
3 Discussion 6
4 Historical Notes 9
2
Relevance Models in Information Retrieval 11
Victor Lavrenko and W Bruce Croft
1 Introduction 11
2 Relevance Models 15
3 Estimating a Relevance Model 18
4 Experimental Results 31
5 Conclusions 51
6 Acknowledgments 54
3
Language Modeling and Relevance 57
Karen SpGlrck lones, Stephen Robertson, Djoerd Hiemstra and Hugo Zaragoza
1 Introduction 58
2 Relevance in LM 59
3 A possible LM approach: parsimonious models 65
4 Concluding comment 70
4
Contributions of Language Modeling to the Theory and Practice of IR 73
Warren R. Greift and William T. Morgan
1 Introduction 73
2 What is Language Modeling in IR 76
3 Simulation Studies of Variance Reduction 81
4 Continued Exploration 89
v
vi LANGUAGE MODELING FOR INFORMATION RETRIEVAL
5
Language Models for Topic Tracking 95
Wessei Kraaij and Martijn Spitters
1 Introduction 95
2 Language models for IR tasks 96
3 Experiments 107
4 Discussion 116
5 Conc1usions 120
6
A Probabilistic Approach to Term Translation for Cross-Lingual Retrieval 125
Jinxi Xu and Ralph Weischedei
1 Introduction 125
2 A Probabilistic Model for CLIR 127
3 Estimating Term Translation Probabilities 129
4 Related Work 130
5 Test Collections 131
6 Comparing CLIR with Monolingual Baseline 132
7 Comparing Probabilistic and Structural Translations 132
8 Comparing Probabilistic Translation and MT 134
9 Measuring CLIR Performance as a Function of Resource Sizes 135
10 Reducing the Translation Cost of Creating a Parallel Corpus 137
11 Conc1usions 139
7
Using Compression-Based Language Models for Text Categorization 141
William 1. Teahan and David J. Harper
1 Background 142
2 Compression models 143
3 Bayes Classifiers 144
4 PPM-based language models 146
5 Experimental results 147
6 Discussion 160
8
Applications of Score Distributions in Information Retrieval 167
R. Manmatha
1 Introduction 167
2 Related Work 169
3 Modeling Score Distributions of Search Engines 171
4 Combining Search Engines Indexing the Same Database 178
5 Applications to Filtering and Topic Detection and Tracking 183
6 Combining Search Engines Indexing Different Databases or Differ-
ent Languages 185
7 Conc1usion 186
8 Acknowledgements 186
9
An Unbiased Generative Model for Setting Dissemination Thresholds 189
Yi Zhang and Jamie Callan
1 Introduction 189
2 Generative Models of Dissemination Thresholds 191
3 The Non-Random Sampling Problem & Solution 196
4 Experimental Methodology 201
Contents vii
5 Experimental Results 204
6 Conclusion 213
10
Language Modeling Experiments in Non-Extractive Summarization 219
Vibhu O. Mittal, Michael J. Witbrock
1 Introduction 220
2 Related Work 221
3 Statistical Models of Gisting 223
4 Training the Models 226
5 Output and Evaluation 231
6 Conclusion 241
Index 245
Preface
A statisticallanguage model, or more simply a language model, is a prob
abilistic mechanism for generating text. Such adefinition is general enough
to include an endless variety of schemes. However, a distinction should be
made between generative models, which can in principle be used to synthesize
artificial text, and discriminative techniques to classify text into predefined cat
egories.
The first statisticallanguage modeler was Claude Shannon. In exploring the
application of his newly founded theory of information to human language,
Shannon considered language as a statistical source, and measured how weH
simple n-gram models predicted or, equivalently, compressed natural text. To
do this, he estimated the entropy of English through experiments with human
subjects, and also estimated the cross-entropy of the n-gram models on natural
text.1 The ability of language models to be quantitatively evaluated in tbis way
is one of their important virtues.
Of course, estimating the true entropy of language is an elusive goal, aiming
at many moving targets, since language is so varied and evolves so quickly. Yet
fifty years after Shannon's study, language models remain, by all measures, far
from the Shannon entropy liInit in terms of their predictive power. However,
tbis has not kept them from being useful for a variety of text processing tasks,
and moreover can be viewed as encouragement that there is still great room for
improvement in statisticallanguage modeling.
In the past several years a new framework for information retrieval has
emerged that is based on statisticallanguage modeling. The approach differs
from traditional probabilistic approaches in interesting and subtle ways, and
is fundamentally different from vector space methods. It is striking that the
language modeling approach to information retrieval was not proposed until
the late 1990s; however, until recently the information retrieval and language
modeling research communities were somewhat isolated. The communities are
'c. E. Shannon. Prediction and entropy of printed English. Bell System Technical Journal, Vol. 30, pp.
51-64,1951.
ix
x LANGUAGE MODELING FOR INFORMATION RETRIEVAL
now beginning to work more c10sely together, and research at a number of sites
has confirmed that the language modeling approach is an effective and theoret
ically attractive probabilistic framework for building IR systems. But there is
still groundwork to do in understanding the basics of this new approach, and
many possibilities exist for further development of the framework.
This book contains a collection of papers that together give an overview of
the recent research activity in the area of language modeling for information
retrieval. The book grew out of a workshop on this topic that took place at
Carnegie Mellon University on May 31 and June 1, 2001, under the support of
ARDA. Several of the presentations were extended into the papers that appear
here.
The papers refiect the major themes of the workshop, and can be thought
of as falling into three broad, overlapping categories. Several papers address
the mathematical formulation and interpretation of the language modeling ap
proach. These papers focus on the important but subtle and elusive concept
of relevance, which was a recurring theme of the workshop. Another category
is the development of the language modeling approach for improving ad hoc
retrieval-the problem of finding documents to answer unstructured queries
on unrestricted topics. In these papers the focus is on the problems of variance
reduction, better estimation of relevance models, improving cross-Ianguage re
trieval, score normalization and the distribution of scores in relevant and non
relevant documents. The third category of papers treats the use of language
modeling methods in other important application areas within information re
trieval, inc1uding topic tracking, text classification, and summarization.
The diversity of the papers collected here is an indication of the richness of
language modeling methods for approaching information retrieval problems.
We believe that the papers provide an interesting cross-section of current work,
and hope that they will inspire future research in this rapidly progressing field.
BRUCECROFf
JOHN LAFFERTY
Contributing Authors
Jamie Callan is an Associate Professor in the School of Computer Science
at Carnegie Mellon University. He earned his Ph.D. from the University of
Massachusetts, Amherst.
Bruce Croft is a Distinguished Professor in the Department of Computer Sci
ence at the University of Massachusetts, Amherst. He received his Ph.D. from
the University of Cambridge.
Warren Greiff is a Principal Scientist at The MITRE Corporation. He earned
his Ph.D from the University of Massachusetts, Amherst.
David Harper is a Research Professor in the School of Computing at The
Robert Gordon University, Aberdeen, Scotland, and Director of the Smart Web
Technologies Center.
Wessei Kraaij is a senior researcher and project manager at the department of
Multimedia Technology and Statistics ofTNO TPD (Netherlands Organization
of Applied Scientific Research).
Djoerd Hiemstra wrote bis Ph.D. thesis on the use of language models for
information retrieval. He is currently an Assistant Professor at the Database
Group of the University of Twente.
John Lafferty is an Associate Professor in the Computer Science Department
at Carnegie Mellon University. He received his Ph.D. from Princeton Univer
sity.
Victor Lavrenko is a Research Assistant in the Department of Computer Sci
ence at the University of Massachusetts, Amherst.
xi